Tsetlin Machine ensemble learns collaboratively without sharing raw data
Decentralized Tsetlin Machines achieve centralized-level accuracy without data exchange
A paradigm for decentralized collaborative learning among an ensemble of Tsetlin Machines using consensus-based inference is proposed. Each agent maintains a private TM model with no raw data exchange. Inference combines individual predictions into a global consensus. The paradigm accommodates heterogeneous TM-based agents with differing data acquisition means, local data distributions, or computational resources. Experiments using two-dimensional grid and connected graph network topologies demonstrate classification accuracies comparable to centralized models.
- Each agent trains a private Tsetlin Machine on a vertical feature partition without sharing raw data.
- Inference combines individual predictions via consensus (e.g., majority vote), matching centralized accuracy.
- Works on grid and connected graph topologies, supporting heterogeneous agents with different data and compute.
Why It Matters
Enables privacy-preserving, decentralized AI with interpretable Tsetlin Machines—ideal for edge computing and multi-sensor fusion.